On Thu, Mar 27, 2014 at 08:11:59PM +0100, Kurt Roeckx wrote:
> On Thu, Mar 27, 2014 at 05:20:06PM +0000, Viktor Dukhovni wrote:
> > What would an O/S distribution do with "SYSTEM" that would make it
> > better than DEFAULT or ALL?
>
> You really do not want to use DEFAULT. And some people even set
> it to ALL having no idea what that does.
Postfix uses ALL, and we know exactly what we're doing and why.
> We either need sane defaults, or some way for applications to
> select sane values without every application having to be updated.
The defaults *are* sane. They are maximally interoperable (excluding
the problems with servers that choke on large client HELLO messages),
and put stronger ciphers ahead of weaker ones.
I think your view is too narrow. TLS is used in many different
applications beyond browsers, and it is rather risky to change the
cipherlist for a set of unknown applications. A flag day change
in the cipher list of all applications, even ones whose needs are
not known would be bad.
I would certainly not use a volatile cipher-list that is changed
at will by the O/S vendor. If this is an administrator choice, it
is more reasonable. Having a named set of such choices in a
system-wide openssl.cnf, would be much better than a single "SYSTEM"
cipherlist. The only difference between "SYSTEM" and "DEFAULT" is
that "DEFAULT" is compiled-in, while "SYSTEM" is in a config file.
If you really want to do what you propose, then perhaps it should
be possible to override "DEFAULT" in openssl.cnf. That's a blunt
tool, the proposal for multiple named cipherlists is saner.
Postfix uses effectively this approach, in that it has internally
an "export", "low", "medium" and "high" grade cipherlist, and users
pick a cipher-grade, not a cipherlist.
--
Viktor.
______________________________________________________________________
OpenSSL Project http://www.openssl.org
Development Mailing List [email protected]
Automated List Manager [email protected]